What will we learn?

• How to determine which explanatory variables affect a model’s prediction for a single observation.

  • Break-down plots
  • Ceteris-paribus profiles
  • Local-model approximations
  • Shapley values;

• Techniques to examine predictive models as a whole.

  • Partial-dependence plots
  • Variable-importance plots

• Charts that can be used to present the key information in a quick way

• Tools and methods for model comparison